CN113780439B - Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation - Google Patents
Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation Download PDFInfo
- Publication number
- CN113780439B CN113780439B CN202111081849.2A CN202111081849A CN113780439B CN 113780439 B CN113780439 B CN 113780439B CN 202111081849 A CN202111081849 A CN 202111081849A CN 113780439 B CN113780439 B CN 113780439B
- Authority
- CN
- China
- Prior art keywords
- satellite data
- domain
- satellite
- module
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation, which adopts an unsupervised domain adaptation method to solve the difficulties of cross-satellite, detection channel difference and the like, extracts domain invariant features based on learning data distribution of two domains, reduces domain difference in a high-level semantic feature space, and further realizes multi-cloud identification on a target domain satellite without labels, and realizes average class precision of various different cloud classes. The application applies the unsupervised domain adaptation to the satellite remote sensing field, carries out cloud classification on certain satellites under the condition of no labeling, and realizes the combination of the remote sensing meteorological field and the deep learning. The system can reduce the data marking work of the new satellite by the meteorological professional, get rid of the dependence of the meteorological on the numerical calculation monitoring mode, promote the new satellite to be rapidly applied to the ground and realize the product feedback after the new satellite is used up, and lay the foundation for the application of more different series of satellites.
Description
Technical Field
The application relates to the technical field of meteorological monitoring, in particular to a multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation.
Background
Satellite technology has been rapidly developed based on the aerospace industry in recent years, and meteorological monitoring based on satellites is more closely related to life of people. However, knowledge among different satellites is difficult to migrate and apply, so that the method has extremely important influence on research on domain adaptation migration among different meteorological satellites, has a pushing effect on development of aerospace industry and meteorological satellites in China, can greatly save time and resources of professional researchers of the meteorological satellites, and has great significance on intensive and careful research of the weather satellites.
The data monitored by the meteorological satellite in space are transmitted back to the earth surface and are classified into different cloud types by means of manual labeling, numerical mode calculation and the like, and the data generally comprise a plurality of channels, such as a visible light channel, a near infrared channel, a thermal infrared channel and the like. The information that professional meteorological personnel gave through the data can divide out more than ten kinds of cloud types through manual annotation or numerical mode operation, and it includes: clear cloud, cloud layer cloud, deep convection cloud, high-altitude layer cloud, rain cloud layer cloud, layer cloud and the like, and inversion, forecast and early warning of weather are carried out through cloud types.
However, different types of meteorological satellites have different observation bands, number of detection channels, etc. For example, the wind cloud No. 4A satellite is a weather satellite which is independently developed in the last 70 th century in China, and has 14 channels in total, including visible light, short wave infrared, medium wave infrared, long wave infrared and other channels. The sunflower satellite No. 8 developed in Japan has a great difference, taking the detection band and the frequency as an example, the sunflower satellite No. 8 has 16 channels in total, and compared with the wind cloud satellite No. 4, the sunflower satellite has two more channels (a B channel and an infrared channel in a visible light channel).
Therefore, there is a huge domain difference between different satellites, and since most of the current deep learning data is based on data driving type, one model can only better adapt to one type of data, so that model migration between two satellites cannot be directly performed, and the existing trained model cannot show better performance.
In the prior art, a cloud class automatic identification method and system as disclosed in CN108846334a comprises a data acquisition module, a cloud image identification module, a cloud image result display module and a model construction module, wherein the cloud image result display module is respectively connected with the data acquisition module, the cloud image identification module and the model construction module through signals, an improved Dense Net is provided on the basis of densely connecting a convolution network (Dense Net), and technologies such as mobile phone APP development and camera monitoring video processing are combined, so that the problems that the types of clouds are various, the pertinence of partially extracted features is high, effective features are difficult to extract from massive cloud image data, and the internal connection between different cloud images cannot be fully excavated are solved. However, this solution has the following drawbacks: (1) tag cannot acquire: for our target domain satellite data, the available tags cannot be obtained, so that the target domain data cannot be learned and trained by using a supervised deep learning method, and the unsupervised deep learning method cannot meet the application requirements; (2) high labor cost: for satellite data without perfecting cloud products, manual labeling is mainly relied on, and particularly, the labeling of the satellite data belongs to pixel-level labeling, so that the labor cost is high, and the labeling period is long; (3) existing models are low generalization on new data: the prior art does not consider the mutual reference of experiences among a plurality of satellites, and the existing trained deep learning model has the problems of poor generalization and the like on new data.
Disclosure of Invention
Aiming at the task of identifying the cloud types, the application aims at solving the problem of how to apply the experience knowledge acquired on a mature satellite (a source domain satellite) to a new satellite (a target domain satellite) without labels, and further provides a cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation.
In order to achieve the above object, the present application provides the following technical solutions:
the application firstly provides a multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation, which is characterized by comprising the following modules:
the input module is used for taking the target domain satellite data and the source domain satellite data as input and inputting the input data into the satellite data preprocessing module;
the satellite data preprocessing module is used for preprocessing satellite data, normalizing pixel levels, dimension processing and processing the lowest spatial resolution of different channels of the same satellite according to the input satellite data;
the source domain satellite data feature extraction and satellite data segmentation module is used for carrying out feature extraction on the satellite data of the source domain and semantic segmentation on the source domain data and the target domain data;
the target domain satellite data feature extraction module is used for extracting features of satellite data of a target domain;
the target domain satellite data domain discriminator module is used for discriminating and feeding back high-dimensional semantic features of satellite data generated by the source domain and the target domain, and the high-dimensional semantic features are in an antagonistic relation with the target domain satellite data feature extraction module;
and the output module is used for outputting the marked file form of the target domain satellite.
Further, the input module comprises a source domain data input module and a target domain data input module, wherein the target domain data input module and the source domain data input module respectively take target domain satellite data and source domain satellite data as inputs, and jointly input the target domain satellite data and the source domain satellite data into the satellite data preprocessing module for operation.
Further, in the satellite data preprocessing module, the original format of the satellite data is a three-dimensional array, and the satellite data preprocessing comprises unification of resolution and longitude and latitude directions, unification of data formats, normalization processing of pixel layers and unification processing of dimensions.
Further, in the satellite data preprocessing module, the satellite data is preprocessed and then converted into a data format usable by the deep learning framework.
Further, in the satellite data preprocessing module, the dimension processing is unified into the area size of the number of the high×wide×satellite channels.
Further, in the source domain satellite data feature extraction and satellite data segmentation module, feature extraction is performed on the satellite data of the source domain, wherein the feature extraction mainly comprises a deep neural network for feature extraction, which is composed of a plurality of convolution layers, and semantic segmentation is performed on the satellite data of the source domain and the target domain, which mainly comprises a semantic segmentation neural network, which is composed of a plurality of convolution layers and an up-sampling layer, and finally a segmentation result is output.
Further, in the target domain satellite data feature extraction module, feature extraction is performed on the target domain satellite data, and the feature extraction module mainly comprises a deep neural network formed by a plurality of layers of convolution layers and extracts semantic features of the high-dimensional target domain satellite data.
The application also provides a control method of the multi-cloud identification system based on the different types of meteorological satellites adapting to the unsupervised domain, which comprises the following steps:
s1, obtaining labels of two kinds of satellite data and source domain satellite data with enough quantity and transmitting the labels to a satellite data preprocessing module;
s2, transmitting the preprocessed source domain satellite data into a source domain satellite data feature extraction and satellite data segmentation module for pre-training, and obtaining segmentation training results of the source domain data;
s3, transmitting the preprocessed target domain satellite data to a target domain satellite data feature extraction module and a target domain satellite data domain discriminator module, and performing pre-training by combining the segmentation training result in the S2, namely updating only the target domain satellite data domain discriminator module;
and S4, training is continued after the pre-training is completed, namely the target domain satellite data domain discriminator module and the target domain satellite data feature extraction module are updated in an iterative mode, and a final prediction result is output.
Compared with the prior art, the application has the beneficial effects that:
according to the multi-cloud identification system of different types of meteorological satellites based on the unsupervised domain adaptation, the unsupervised domain adaptation method is adopted, the difficulties of cross-satellite, detection channel difference and the like are solved, domain invariant features are extracted based on learning data distribution of two fields, domain difference is reduced in a high-level semantic feature space, multi-cloud identification is further achieved on a target domain satellite without labels, and average class precision of various different cloud classes is achieved.
The application applies the unsupervised domain adaptation to the satellite remote sensing field, carries out cloud classification on certain satellites under the condition of no labeling, and realizes the combination of the remote sensing meteorological field and the deep learning.
The system can reduce the data marking work of the new satellite by the meteorological professional, get rid of the dependence of the meteorological on the numerical calculation monitoring mode, promote the new satellite to be rapidly applied to the ground and realize the product feedback after the new satellite is used up, and lay the foundation for the application of more different series of satellites.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings for a person having ordinary skill in the art.
Fig. 1 is a schematic diagram of a multi-cloud identification system based on different types of meteorological satellites adapted by an unsupervised domain according to an embodiment of the present application.
Fig. 2 is a flowchart of a control method of a multi-cloud identification system based on different types of meteorological satellites adapted by an unsupervised domain according to an embodiment of the present application.
Detailed Description
For a better understanding of the present technical solution, the method of the present application is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the application provides a multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation, which comprises the following modules:
(1) Input module
And taking the target domain satellite data and the source domain satellite data as inputs, and inputting the inputs to a subsequent module for operation.
(2) Satellite data preprocessing module
The method mainly comprises the steps of preprocessing satellite data according to input satellite data, wherein the preprocessing mainly comprises resolution, unification of longitude and latitude directions and unification of data formats (processing is a data format usable by a deep learning framework), normalization processing of pixel layers, dimension processing (unification is h x w x c area size), and processing of the lowest spatial resolution of different channels of the same satellite.
Specifically, the satellite data is input into the satellite data preprocessing module, the original format of the satellite data is a three-dimensional array, and the satellite data is exemplified by sunflower number 8 satellite data of the strait of Zhejiang, which comprises 16 channels of satellite data, and is labeled as labeled data with the same size of 1 channel. The minimum spatial resolution processing of different channels of the same satellite refers to the area normalization processing of the area represented by one pixel of different channels of the same satellite, and the data format processing refers to the unification of the data formats of different satellites, so that the data formats of different satellites are conveniently input into a deep learning model. The satellite data preprocessing module mainly aligns the data of the source domain and the target domain on the spatial resolution, and transmits the preprocessed data as output to the subsequent module. Thereby ensuring the uniformity, the integrity and the accuracy of the data in the subsequent data training process.
(3) Source domain satellite data feature extraction and satellite data segmentation module
The method mainly comprises the steps of extracting features of satellite data of a source domain, wherein the feature extraction mainly comprises a deep neural network for feature extraction, which is composed of a plurality of convolution layers, and semantic segmentation of the satellite data of the source domain and a target domain mainly comprises a semantic segmentation deep neural network, which is composed of a plurality of convolution layers and an up-sampling layer, and finally outputting segmentation results.
Specifically, in the source domain satellite data feature extraction and satellite data segmentation module, the main functions of the source domain satellite data feature extraction and semantic segmentation of source domain data and target domain data are as follows. Its input is the preprocessed source domain satellite data and labels. In a specific structure, the first layer convolution module mainly aims to eliminate the influence caused by the channel difference. For source domain data, the module extracts deep semantic information from the multi-layer residual error network module and restores the image through up-sampling. For the target domain data, the target domain satellite data feature extraction module inputs the output of the target domain satellite data feature extraction module in an intermediate stage, and performs partial feature extraction and up-sampling.
(4) Target domain satellite data feature extraction module
The feature extraction is carried out on the satellite data of the target domain, and the feature extraction mainly comprises a deep neural network formed by a plurality of convolution layers, so that the semantic features of the satellite data of the target domain with high dimension are extracted.
Specifically, in the target domain satellite data feature extraction module, the main function of the feature extraction module is to extract features of the target domain satellite data. Its input is target domain satellite data. The first layer convolution module is mainly used for eliminating influence caused by channel difference. The method comprises the steps of extracting features of satellite data in a target domain to obtain deep semantic information, and transmitting the deep semantic information back to a source domain satellite data feature extracting and satellite data dividing module. Which acts as a party in the antagonizing network to interact with the target-domain satellite data discriminator module.
(5) Target domain satellite data domain discriminator module
And judging the high-dimensional semantic features of the satellite data generated by the source domain and the target domain, and feeding back the high-dimensional semantic features to be in an antagonistic relation with the satellite data feature extraction module of the target domain.
Specifically, in the target domain satellite data domain discriminator module, its main role is to discriminate whether the sample is from the source domain (actual distribution) or the target domain (learning distribution). Its inputs come from source domain satellite data feature extraction and semantic features of different depths of the satellite data segmentation module. In the iterative learning process, if the identifier module cannot separate the target domain features from the source domain features in the iterative process, the feature extractor module extracts features more similar to the source domain and the target domain. In the countermeasure training, domain invariant features are extracted by learning the data distribution of two domains, and domain differences are reduced in a high-level semantic feature space, so that unsupervised domain adaptation is realized. In addition to using a domain discriminator module, an optimization model architecture such as a tag discriminator can be inserted.
(6) Output module
And outputting the file form which is marked by the target domain satellite.
The application also provides a control method for the multi-cloud identification system based on different types of meteorological satellites adapting to an unsupervised domain, as shown in fig. 2, the method comprises the following steps:
s1: obtaining labels of two kinds of satellite data and source domain satellite data in a sufficient quantity and transmitting the labels to a satellite data preprocessing module;
s2: the preprocessed source domain satellite data are transmitted to a source domain satellite data feature extraction and satellite data segmentation module for pre-training, and segmentation training results of the source domain data are obtained;
s3: transmitting the preprocessed target domain satellite data into a target domain satellite data feature extraction module and a target domain satellite data domain discriminator module, and performing pre-training by combining the segmentation training result in the step S2, namely updating only the target domain satellite data domain discriminator module;
s4: and after the pre-training is finished, training is continued, namely the target domain satellite data domain discriminator module and the target domain satellite data feature extraction module are updated in an iterative mode, and a final prediction result is output.
Examples
The sunflower 8 satellite data in Japan is used as a source domain, the Chinese domestic Fengyun four-satellite A satellite is used as a target domain, and the transfer learning of unsupervised domain adaptation is realized, so that the label-free labeling of the Fengyun four-satellite A satellite is realized.
In the data preprocessing module, two types of three-dimensional satellite data are input, the processing comprises the unification of resolution, the unification of longitude and latitude directions near the strait of Zhejiang, the unification of data format into the tfreeord format of tensorsurface, the normalization processing of pixel level and the unification processing of dimension. The final preprocessed sunflower number 8 satellite data is a 512 x 16 three-dimensional array, the sunflower number 8 satellite tag is a 512 x 1 three-dimensional array, the Fengyun four-satellite A satellite data is a 512 x 14 three-dimensional array, and channel difference exists between the two satellites.
Then, the preprocessed source domain data is input into a source domain satellite data feature extraction and satellite data segmentation module for pre-training, which is similar to a semantic segmentation algorithm, mainly learns the distribution of the source domain data, and obtains a sunflower number 8 satellite data segmentation training result.
After the pre-training is finished, the sunflower satellite data No. 8 is input into a satellite data feature extraction and satellite data segmentation module, and the Fengyun four-satellite A satellite data is input into a target domain satellite data feature extraction module. Iterative updating based on countermeasure training to generate a countermeasure network GAN is performed by a target domain satellite data domain discriminator module, the domain gap is reduced, and cloud class distribution on target domain data is learned.
And finally, outputting cloud classification semantic segmentation results near Zhejiang strait A type by the output module, wherein different colors represent different kinds of cloud distribution.
The application mainly relates to an unsupervised domain adaptive transfer learning mode, and belongs to the technical field of deep learning. The deep learning is realized by simulating a neural network of the human brain, building different convolution layers, pooling layers and the like to learn the internal rules and the representation layers of sample data, and finally realizing the learning and the discrimination of the data.
The unsupervised domain adaptation is a kind of transfer learning, and is to extract the domain invariant features by generating modes such as countermeasure learning and the like under the condition that the target domain does not have labels, and reduce the domain difference between the source domain and the target domain in a high-level semantic feature space, so as to realize the unsupervised learning on the target domain.
The application discloses a multi-cloud identification system based on different types of meteorological satellites adapting to an unsupervised domain and a control method thereof, which have the following advantages compared with the prior art:
(1) The labor cost is low: according to the application, intelligent labeling can be performed on satellite data without labeling in an unsupervised learning and deep learning mode, so that the workload of meteorological staff is reduced, and the labor cost is reduced.
(2) The application period is short: compared with the manual annotation time, the method has the advantages that the cloud distribution on the target domain can be quickly learned after model training is completed, and the application period is shorter.
(3) The universality is strong: the application can be conveniently transplanted to the transfer learning between any two satellites, has strong universality, is simple and convenient to apply and assists in manual identification.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may be modified or some technical features may be replaced with others, which may not depart from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims (8)
1. A multi-cloud identification system for different types of meteorological satellites based on unsupervised domain adaptation, comprising the following modules:
the input module is used for taking the target domain satellite data and the source domain satellite data as input and inputting the input data into the satellite data preprocessing module;
the satellite data preprocessing module is used for preprocessing satellite data, normalizing pixel levels, dimension processing and processing the lowest spatial resolution of different channels of the same satellite according to the input satellite data;
the source domain satellite data feature extraction and satellite data segmentation module is used for carrying out feature extraction on the satellite data of the source domain and semantic segmentation on the source domain data and the target domain data;
the target domain satellite data feature extraction module is used for extracting features of satellite data of a target domain;
the target domain satellite data domain discriminator module is used for discriminating and feeding back high-dimensional semantic features of satellite data generated by the source domain and the target domain, and the high-dimensional semantic features are in an antagonistic relation with the target domain satellite data feature extraction module;
and the output module is used for outputting the marked file form of the target domain satellite.
2. The multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein the input module comprises a source domain data input module and a target domain data input module, and the target domain data input module and the source domain data input module respectively take target domain satellite data and source domain satellite data as inputs, and jointly input the target domain satellite data and the source domain satellite data into the satellite data preprocessing module for operation.
3. The multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein in the satellite data preprocessing module, the original format of satellite data is a three-dimensional array, and the satellite data preprocessing comprises unification of resolution and longitude and latitude directions, unification of data formats and normalization processing of pixel levels.
4. The multi-cloud identification system for different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein in the satellite data preprocessing module, the satellite data is preprocessed and then converted into a data format usable by a deep learning framework.
5. The multi-cloud identification system for different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein in the satellite data preprocessing module, the dimension processing is unified into the area size of the number of high x wide x satellite channels.
6. The multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein in the source domain satellite data feature extraction and satellite data segmentation module, feature extraction is performed on the satellite data of the source domain, wherein the feature extraction mainly comprises a deep neural network for feature extraction composed of multiple convolution layers, semantic segmentation is performed on the satellite data of the source domain and the target domain mainly comprises a semantic segmentation neural network composed of multiple convolution layers and an up-sampling layer, and finally a segmentation result is output.
7. The multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation according to claim 1, wherein in the target domain satellite data feature extraction module, feature extraction is performed on satellite data of a target domain, and the system mainly comprises a deep neural network composed of multiple layers of convolution layers, so that semantic features of high-dimensional target domain satellite data are extracted.
8. The method for controlling a multi-cloud identification system for different types of meteorological satellites based on unsupervised domain adaptation according to any one of claims 1 to 7, comprising the steps of:
s1, obtaining labels of two kinds of satellite data and source domain satellite data with enough quantity and transmitting the labels to a satellite data preprocessing module;
s2, transmitting the preprocessed source domain satellite data into a source domain satellite data feature extraction and satellite data segmentation module for pre-training, and obtaining segmentation training results of the source domain data;
s3, transmitting the preprocessed target domain satellite data to a target domain satellite data feature extraction module and a target domain satellite data domain discriminator module, and performing pre-training by combining the segmentation training result in the S2, namely updating only the target domain satellite data domain discriminator module;
and S4, training is continued after the pre-training is completed, namely the target domain satellite data domain discriminator module and the target domain satellite data feature extraction module are updated in an iterative mode, and a final prediction result is output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111081849.2A CN113780439B (en) | 2021-09-15 | 2021-09-15 | Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111081849.2A CN113780439B (en) | 2021-09-15 | 2021-09-15 | Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113780439A CN113780439A (en) | 2021-12-10 |
| CN113780439B true CN113780439B (en) | 2023-09-22 |
Family
ID=78844034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111081849.2A Active CN113780439B (en) | 2021-09-15 | 2021-09-15 | Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113780439B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062753A (en) * | 2017-12-29 | 2018-05-22 | 重庆理工大学 | The adaptive brain tumor semantic segmentation method in unsupervised domain based on depth confrontation study |
| KR101855652B1 (en) * | 2018-02-13 | 2018-06-25 | 대한민국 | System of total cloud cover retrieval from satellite observation using machine learning and method thereof |
| CN113052201A (en) * | 2020-12-22 | 2021-06-29 | 中国人民解放军61540部队 | Satellite cloud picture cloud classification method based on deep learning |
-
2021
- 2021-09-15 CN CN202111081849.2A patent/CN113780439B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108062753A (en) * | 2017-12-29 | 2018-05-22 | 重庆理工大学 | The adaptive brain tumor semantic segmentation method in unsupervised domain based on depth confrontation study |
| KR101855652B1 (en) * | 2018-02-13 | 2018-06-25 | 대한민국 | System of total cloud cover retrieval from satellite observation using machine learning and method thereof |
| CN113052201A (en) * | 2020-12-22 | 2021-06-29 | 中国人民解放军61540部队 | Satellite cloud picture cloud classification method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113780439A (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Du et al. | Pavement distress detection and classification based on YOLO network | |
| CN109949316A (en) | A kind of Weakly supervised example dividing method of grid equipment image based on RGB-T fusion | |
| Yuan et al. | Large-scale solar panel mapping from aerial images using deep convolutional networks | |
| CN107133569B (en) | Monitoring video multi-granularity labeling method based on generalized multi-label learning | |
| CN110879989B (en) | Ads-b signal target identification method based on small sample local machine learning model | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN104992223A (en) | Dense population estimation method based on deep learning | |
| CN115131627B (en) | Construction and training method of lightweight plant disease and pest target detection model | |
| CN110675421B (en) | Depth image collaborative segmentation method based on few labeling frames | |
| Chen et al. | Agricultural remote sensing image cultivated land extraction technology based on deep learning | |
| CN108345900B (en) | Pedestrian re-identification method and system based on color texture distribution characteristics | |
| CN113205039B (en) | Power equipment fault image recognition disaster investigation system and method based on multiple DCNN networks | |
| CN111461006B (en) | Optical remote sensing image tower position detection method based on deep migration learning | |
| CN105005852A (en) | Image analysis based intelligent monitoring system for dormitory environment | |
| Yu et al. | Convolutional neural network with feature reconstruction for monitoring mismatched photovoltaic systems | |
| CN111242028A (en) | Remote sensing image ground object segmentation method based on U-Net | |
| CN111126155A (en) | Pedestrian re-identification method for generating confrontation network based on semantic constraint | |
| CN114241423A (en) | Intelligent detection method and system for river floaters | |
| CN111414855B (en) | Telegraph pole sign target detection and identification method based on end-to-end regression model | |
| CN117496426A (en) | Precast beam procedure identification method and device based on mutual learning | |
| CN113780439B (en) | Multi-cloud identification system of different types of meteorological satellites based on unsupervised domain adaptation | |
| CN116205905B (en) | Power distribution network construction safety and quality image detection method and system based on mobile terminal | |
| CN101876993A (en) | Method for extracting and retrieving textural features from ground digital nephograms | |
| Liu et al. | Integration transformer for ground-based cloud image segmentation | |
| CN110826478A (en) | Aerial photography illegal building identification method based on countermeasure network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |